.png){kind=icon}
返回
StriPre: A VM Disk Migration Prefetching Algorithm
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2025.3545836.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Virtual machine (VM) cross-cloud migration involves relocating virtual machine instances between different cloud environments, offering advantages such as improved scalability, disaster recovery capabilities, and efficient resource allocation optimization. Two strategies exist for live migration: pre-copy and post-copy. These methods are initially applied to memory migration, but their drawbacks, such as data redundancy and poor virtual machine IO performance are amplified in cross-cloud disk migration due to high network latency, limited transmission bandwidth, and large data volumes, thereby constraining the performance of cross-cloud virtual machine migration. To reduce the performance impact caused by the latency of cross-cloud migration, we present StriPre in this paper. StriPre is a prefetching algorithm based on disk access and striped feature design, leveraging the alternating characteristics of disk accesses and performing striping recognition on IO requests. Compared to the commonly used sequential prefetching in memory, StriPre's prefetch hit rate has increased by 1.2x to 2.2x. Our experiments have demonstrated that StriPre can effectively reduce the virtual machine startup time during cross-cloud disk migration. Under extreme cross-cloud latency of 160ms, StriPre can shorten the startup time by 3.48x compared to sequential prefetching.
Keyword:
Prefetching
Virtual machines
Costs
Cloud computing
Prediction algorithms
Approximation algorithms
Real-time systems
Kernel
Business
Accuracy
Cross-cloud migration
data prefetch algorithm
cloud computing
